Primal-dual algorithms for discounted Markov decision processes.
Randy CogillPublished in: ECC (2015)
Keyphrases
- markov decision processes
- policy iteration
- primal dual
- factored mdps
- interior point
- state space
- dynamic programming
- reachability analysis
- optimal policy
- finite state
- infinite horizon
- reinforcement learning
- linear program
- interior point methods
- convergence rate
- linear programming
- finite horizon
- transition matrices
- simplex algorithm
- learning algorithm
- reinforcement learning algorithms
- partially observable markov decision processes
- decision theoretic planning
- optimization problems
- semidefinite programming
- markov decision process
- approximation algorithms
- average reward
- worst case
- discounted reward
- convex optimization
- partially observable
- state and action spaces
- action sets
- search algorithm
- machine learning
- stochastic shortest path